BOP – Ellipsometer
Responsible person: Per-Anders Hansen
Mob.: 4824 5746
Office: ØU35
Contents
Doing a measurement .................................................................................... 2
Measurement sequence: ................................................................................ 2
Analysing ....................................................................................................... 2
Transparent films on silicon ........................................................................... 2
Al2O3, La2O3, MgO, SiO2, fully transparent. ..................................................... 2
TiO2 and other UV absorbing materials .......................................................... 3
NiO and other grey mirror-like semitransparent materials............................. 4
Apparently transparent films ......................................................................... 4
Absorbing films on silicon .............................................................................. 4
Transparent substrates .................................................................................. 7
For all non-Si substrates ................................................................................. 8
Checking the validity of your model ............................................................... 9
MSE (Mean Square Error)............................................................................... 9
Refractive index must be physical. ................................................................. 9
Parameter uniqueness ................................................................................. 10
Does it make sense? ..................................................................................... 11
Trouble shooting.......................................................................................... 11
I get an error when trying to measure.......................................................... 11
My green delta curve goes from 0 to infinite ................................................ 11
Risk assessment ........................................................................................... 11
Doing a measurement
Measurement sequence:
1. Turn on the ellipsometer
2. Open CompleteEASE.
3. Put your sample in place. You can see where the beam hits your sample
by holding a piece of paper in front. It anyway hits directly between the
two small holes. And is shaped like a 2x4 mm ellipse.
4. In CompleteEASE, select 70°, Mode: Standard, Medium speed. Then
press “Measure”. If you get an error, see the Troubleshooting section.
5. If you need multiangle measurements, select 65, 70 and 75° before
pushing “Measure”. If you need transmission measurements, talk to me.
6. Save your file and turn of the ellipsometer to save the light bulb.
7. To analyse: Go back to your office, open CompleteEase and open the
analysis tab. When done analysing, the MSE (Mean Square Error) of the
fit should be between 1 and 3. It will be a bit higher for thicker films (>60
nm) and also a bit higher for multiangle measurements.
Analysing
This BOP is not meant as a complete guide on how to analyse any and all
samples, but rather a quick guide to get started. You should read the manual
(by pressing F1 in CompleteEASE or going to the Options tab and opening the
manual) and the pdf’s that are relevant to your type of sample. I can help you
figure out what is relevant and not.
Transparent films on silicon
Transparent films are anything that looks transparent to us. That is, Al2O3, TiO2,
MgO, La2O3 etc.
Al2O3, La2O3, MgO, SiO2, fully transparent.
In the analysis window, press “open” under model.
Go to “Basic” and select “Si with transparent film.mod”. This is a simple Cauchy
function. It only models a thickness and a refractive index, fitted by 3
parameters. Press “Fit”.
That’s it, assuming that you have a nice, flat, even film. If you think your film is
rough, right-click on “Include Surface Roughness = OFF”. This is not a physical
roughness like what AFM will give you. This is just some mathematical
approximation to what roughness will do to the polarized beam.
TiO2 and other UV absorbing materials
For TiO2 and other materials that absorb UV, you may need to add another
function to the Cauchy function. Left-click on the “+” next to “Layer # 1”, and
right-click on “K Amplitude” and “Exponent”.
This allows an absorbtion tail to extend into the measurement range. Even if
the film doesn’t absorb in the measurement range, an absorbtion just outside
the range will still affect the refractive index. Press “Fit”. Now you are done.
NiO and other grey mirror-like semitransparent materials
Some other materials may be fittet with this simple Cauchy approach as well.
As long as the optical absorption of the film follows a nice exponential form
(i.e. increases steadily towards the UV), the TiO2-approach can work. You may
need to select a smaller range, for example 700-900 nm, if this approach
doesn’t fit properly. If a smaller range doesn’t help, you need a different
approach. Anyway, for these types of films it is really important that you check
that the model makes physical sense, as described later.
Apparently transparent films
Some films can look transparent, but the Cauchy function will not give a
satisfactory fit. This means that the model isn’t representative for your sample.
The Cauchy function is representative for transparent films, with the exception
of strongly IR-absorbing films like highly doped ZnO. Therefore, if Cauchy
doesn’t fit properly, that means that your film either is absorbing, but too
weakly for you to notice or it is not really flat or homogeneus (grading, surface
oxidation etc.). Try to fit smaller sections of the spectra. Sometimes the film
can be transparent in the blue region while other times in the red region. If you
find a region where the Cauchy function fits, then that region is transparent
(Cauchy will not be able to fit an absorbing region).
Absorbing films on silicon
Absorbing films are more tricky, and sometimes needs to be solved on an
individual basis. There is however one standard attempt you can try. First, do
the measurement with all angles (except 90°). This will save you time later,
when you figure out you had to do it. More angles gives the models more data,
which reduces the chance that you end up with many different, equally fitting
answers.
Select the “Si with transparent film.mod” model. Select a wavelength range
that you think is more or less transparent. Usually, most absorptions lie in the
visible range, so the IR range is usually safe to assume is transparent. But you
know your material best!
Fit the smaller, selected range with “Si with transparent film.mod”. If you only
want to know the thickness of the film, this is all you need and you can be
happy with the thickness you have. But you should be able to fit everything
and get n and k for all wavelengths. So let’s do that. Now comes the tricky part.
Right click “Cauchy film” and select “Parameterize layer”.
Press “Open Material File”, select “Basic” and then “B-Spline”. Change
“Resolution (eV)” to 0.05. It should say 4 Pts or more directly to the right. If it is
less, you need to select a large range. Press “Fit” and then “Replace Layer”.
When you are back to the normal window, press “Fit” again to fit the reduced
data range with the new B-Spline model.
To go back to the normal data graph window, select “Graph Type” in the top
left of the graph, and select “Psi”. In the model window, expand your layer by
pushing the red + and turn on “Show Advanced Options”, then “Assume
Transparent Region” at the bottom, turn off “Assume Band Gap”, then finally
select “Transparent Region” and type in the smallest value of the range you
select. The large value can be 10000. So if you selected approximatly 750 – 900
nm like me, then type in 750 and 10000 here.
Almost done. Expand “OTHER Options” and change “Increment (eV)” to 0.05
(the same as the resolution earlier). Right click the graph window and select
“Zoom all”. Then left-click “Wvl. Range Expansion Fit” and watch the fun (or
horror).
Final thing to do is to turn off “Fit Opt. Const.”, turn on thickness and
roughness and press “Fit” one last time. This it to optimise thickness and
roughness, which shouldn’t change by more than 1 % (prefferably less).
After doing this, the model should be stable. That means that nothing changes
if you click “Fit” several times. It is really important that you check the validity
of the model after doing something with the B-Spline function! When
everything is ok, you now have the film thickness, refractive index and
absorption coefficient for the whole range. n(λ) and k(λ) are correct, but you
cannot assign much physical reason behind absorption peaks yet. To do that
you need to convert this model (again) from B-Spline into Gen.Osc..
The general oscillator functions are real physical absorption mechanisms which
can give you really good and detailed information about the electronic
transistions in your film. But they can also be quite difficult to handle. Before
starting to play with the Gen.Osc. functions, you should read through the pdf’s
on absorbing films and Gen.Osc. in the ellipsometer folder in the ALD folder.
You should also come talk to me as you can very easily interpret the data
wrong.
Transparent substrates
The big (very big) problem here is that the detector also receives the reflection
coming from the back surface of the substrate, which is depolarized and ruins
everything. There are two ways to deal with this. One that works and one that
doesn’t. The one that doesn’t seem to work well is to turn on “Include backside reflection”, but I have never managed to do a proper fit this way. The way
that do work, and works very well, is to simply tape a piece of diffuse (whiteish) tape on the backside of the glass. Tape has about the same refractive index
as glass, so the light is no longer reflected at the glass surface, but instead
travels into the tape where it gets scattered and no longer hits the detector.
For all non-Si substrates
If you use a non-silicon sibstrate, like quartz or a metal, I strongly advice you to
first measure an empty substrate. Do this as you would measure and model a
sample, except for the model choose “Blank” and click on “none” after
“Substrate:” to insert a fitting function for the substrate.
When you are happy with your model, right click on your substrate and press
“Save Layer Optical Constants” and select either parameterized or tabulated. I
suggest parameterized. From now on, you can select this material (.mat) file
instead of for example “Cauchy”.
Checking the validity of your model
MSE (Mean Square Error)
The MSE is an estimate on how well the model kurve fit the data kurve. For flat
homogeneous transparent film below 100 nm, it should be between 1 and 3.
For thicker films or rougher, more absorbing, layered etc etc., you will have to
accept a higher value. You should look at the kurves and see if your model
seems to follow the curvatures of the data.
Refractive index must be physical.
You should always check that the refractive index is physical (increasing
towards the UV), even for the simples models and the most perfect fits. Below
is a physical and unphysical refractive index. The reason why the right one is
unphysical is that n(λ) can only go down if k(λ) is going up. n(λ) will also start to
turn back up again when k(λ) peaks as seen in the bottom picture. If this is not
the case then the model is unphysical / not correct / not describing your
sample and the results cannot be trusted.
Parameter uniqueness
In the model window, expand (+) “OTHER Options” and click “”Fit Parameter
Uniqueness”. Select the parameter you want to chech, mostly the thickness,
select the range and number of steps and press ok. For Cauchy fits, several
hundred steps can be calculated in a matter of seconds. For more complicated
function it will go slower.
The reulsting graph should look something like this:
Here it is clear that the model is really really happy with 260 nm, and so should
you. Some times you will see several local minimas and the one you ended up
with may not be lowest one. Then you need to change the starting values of
your model. Other time you may experience that the curve is flat at, say, MSE =
6 after the minima. In that case you have a ununique result. Measuring at
multiple angles may help, or do changes in the model. Sometimes variables
gets correlated, for example that changing the density, n(λ) and thickness at
the same time gives many possible answers that fit equally well. This could also
be solved by fixing one of the values (de-clicking it as a fitting parameter).
Does it make sense?
You have to be the final judge of wether or not the model makes sense. You
may get a perfect fit with a model of B-Spline function on top of a quartz
substrate showing clear yellow absorption, but if your sample looks
transparent and the substrate is silicon then the model doesn’t make sense.
Trouble shooting
I get an error when trying to measure
If the program has been on for a long time, or the program was on before the
ellipsometer was turned on, you will get an error message when trying to
measure. Go to the “Hardware” tab and press “Re-initialize”.
My green delta curve goes from 0 to infinite
No it doesn’t, it goes from 0° to 360°. The delta curve is a function of the phase
difference of the detected light. 0° and 360° are both in phase so it just
rescales it.
Risk assessment
There is not risks regarding the use of this instrument, except looking directly
into the light source.
Be careful when moving the detector and light arms. Rough handling may
damage the equipment.